Skip to content

[PyTorch] Make grouped weights opt-in #2678

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ksivaman merged 2 commits into from
Feb 13, 2026
+22 −3
Merged

[PyTorch] Make grouped weights opt-in #2678

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f009c89
Make grouped weights opt-in
ksivaman Feb 13, 2026
197f2e6
Change varname
ksivaman Feb 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 18 additions & 2 deletions tests/pytorch/test_sanity.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -585,10 +585,19 @@ def test_sanity_linear_with_zero_tokens(dtype, bs, model, fp8_recipe, fp8_model_
@pytest.mark.parametrize("fp8_recipe", fp8_recipes)
@pytest.mark.parametrize("fp8_model_params", all_boolean)
@pytest.mark.parametrize("use_bias", all_boolean)
@pytest.mark.parametrize("single_param", all_boolean)
@pytest.mark.parametrize("empty_split", ["first", "last", "middle"])
@pytest.mark.parametrize("num_gemms", [4])
def test_sanity_grouped_linear(
dtype, bs, model, fp8_recipe, fp8_model_params, use_bias, num_gemms, empty_split
dtype,
bs,
model,
fp8_recipe,
fp8_model_params,
use_bias,
single_param,
num_gemms,
empty_split,
):
if NVTE_TEST_NVINSPECT_ENABLED and fp8_model_params:
pytest.skip("FP8 model parameters are not supported in debug mode.")
Expand All @@ -598,6 +607,9 @@ def test_sanity_grouped_linear(
bs = bs * 16
num_tokens = bs * config.max_seqlen_q * (num_gemms - 1)

if single_param:
os.environ["NVTE_ALLOC_CONTIGUOUS_GROUPED_LINEAR_WEIGHTS"] = "1"

if fp8_recipe is not None:
if not is_fp8_supported(config):
pytest.skip("Model config does not support FP8")
Expand All @@ -617,7 +629,8 @@ def test_sanity_grouped_linear(
# Verify that weights are stored in contiguous GroupedTensor storage.
weights = [getattr(te_grouped_linear, f"weight{i}") for i in range(num_gemms)]
if fp8_recipe is None or not (fp8_recipe.delayed() or fp8_recipe.float8_current_scaling()):
check_grouped_tensor_pointers(weights, fp8_recipe)
if single_param:
check_grouped_tensor_pointers(weights, fp8_recipe)

inp_hidden_states = torch.randn(
num_tokens, config.hidden_size, dtype=dtype, requires_grad=True
Expand All @@ -636,6 +649,9 @@ def test_sanity_grouped_linear(
loss.backward()
assert out.shape == (num_tokens, ffn_hidden_size)

if single_param:
del os.environ["NVTE_ALLOC_CONTIGUOUS_GROUPED_LINEAR_WEIGHTS"]


@pytest.mark.parametrize("dtype", param_types)
@pytest.mark.parametrize("fp8_recipe", fp8_recipes)
Expand Down
5 changes: 4 additions & 1 deletion transformer_engine/pytorch/module/grouped_linear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@
from typing import Union, Optional, Callable, Tuple, List
from itertools import chain
import warnings
import os

import functools
import torch
Expand Down Expand Up @@ -793,7 +794,9 @@ def make_grouped_weights(self, defer_init=False) -> None:

def reset_parameters(self, defer_init=False):
super().reset_parameters(defer_init=defer_init)
self.make_grouped_weights(defer_init=defer_init)
# Grouped tensor weights is an opt-in feature.
if bool(int(os.getenv("NVTE_ALLOC_CONTIGUOUS_GROUPED_LINEAR_WEIGHTS", "0"))):
Copy link
Contributor

@greptile-apps greptile-apps bot Feb 13, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This parsing will raise ValueError if the envvar is set to non-numeric values like "true", "false", or empty string. Consider more robust parsing if users might set this directly:

Suggested change
if bool(int(os.getenv("NVTE_ALLOC_CONTIGUOUS_GROUPED_LINEAR_WEIGHTS", "0"))):
if os.getenv("NVTE_ALLOC_CONTIGUOUS_GROUPED_LINEAR_WEIGHTS", "0").lower() in ("1", "true", "yes"):

self.make_grouped_weights(defer_init=defer_init)

def set_tensor_parallel_attributes(self, defer_init=False) -> None:
"""Set attributes needed for TP"""
Expand Down
Loading